Chapter 6 Ants as Mutualists

Joshua Ness, Kailen Mooney, and Lori Lach

6.1 Introduction et al. 2002, Stadler and Dixon 2005, Way 1963; The historical emphasis on the ecological and evolu- seed dispersal in Giladi 2006; ant–plant symbioses tionary importance of antagonistic interactions such in Davidson and McKey 1993; Heil and McKey as competition, predation, and parasitism is increas- 2003; and ant–fungi–bacteria in Poulsen and Currie ingly informed by a recognition of facilitative and 2006) and are featured in several books (e.g. Beattie mutualistic interactions where one or both partici- 1985; Huxley 1991; Rico-Gray and Oliveira 2007; pants receive a net benefit (Bertness and Callaway Stadler and Dixon 2008). We encourage readers to 1994; Bruno et al. 2003; Grosholz 2005; Stachowicz seek out these more in-depth works. Second, the 2001). Interactions between ants and their partners mutualisms we describe often include currencies provide some of the best examples of the reciprocal- based on antagonistic interactions and/or access ly beneficial interactions (Bronstein 1998) and, in to food. Competition, predation, and parasitism of particular, the mutualisms that play critical roles in (and by) ants are treated in other chapters (see structuring community composition and function- Chapters 5, 12, 9, 10, and 11, respectively, and Box ing (e.g. Christian 2001; Kaplan and Eubanks 2005; 6.1), and aspects of ant diet and shelter are the focus Mooney 2007; O’Dowd et al. 2003; Poulsen and Cur- of Chapter 7. In many cases, dissecting mutualistic rie 2006; Wimp and Whitham 2001). Interactions interactions requires an understanding of those cur- between ants and their partners date to 45–60 Mya rencies. (Poulsen and Currie 2006; Stadler and Dixon 2005) We begin by describing mutualisms on the basis and are critical to understanding the evolution and of the resources and services being traded. We ecological success of ants as a taxon. The rewards focus on trophobiotic interactions (Section 6.2), provided by mutualists can increase the survival wherein ants receive access to food resources in and reproduction of ants and colonies, provide the exchange for services provided to the reward pro- fuel that allows ants to collect new resources and ducer (whether plant or insect; bacterial endosym- engage in aggressive behaviours (Davidson 1998), bionts are discussed in Chapter 7), interactions and encourage colonies to reallocate resources to- where ants receive nutritive profit while dispersing wards particular responsibilities and/or locations. plant propagules (seeds and pollen) (Section 6.3), Here, we describe the currencies and dynamics of and the tripartite mutualism among ants, fungal these mutualistic interactions, and highlight recent cultivars, and bacteria, in which food, protection, developments in our understanding of ants’ partici- and dispersal are the currencies (Section 6.4). In pation in mutualisms. each case, we identify instances in which these The complexity and breadth of this topic warrant interactions can have consequences for the larger two caveats. First, the dynamics of particular ant biotic communities and identify characteristics of mutualisms have been the focus of substantive re- ants that make them particularly well suited for views (e.g. refer to plant protection in Bronstein participation in the interaction. We then take a syn- 1998, Heil and McKey 2003; insect tending in Pierce thetic approach to explore elements of context

97 98 ANT ECOLOGY

Box 6.1 ‘Berry’ ants: an eye-popping symbiosis from the rainforest canopy Stephen P. Yanoviak

Successful transmission to a terminal host is the red gasters for ripe fruit (Yanoviak et al. one of the biggest challenges in a parasite’s life 2008b). Unlike the examples mentioned earlier, cycle. Consequently, parasites have evolved a ants are the final hosts for this parasite, and variety of mechanisms to change the behaviour birds function as paratenic hosts (i.e. animals and appearance of intermediate hosts to facil- that transmit parasites to new hosts without itate their consumption by the subsequent becoming infected themselves; Moore 2002). hosts. Several remarkable examples of this Unfortunately, direct evidence for bird preda- phenomenon involve ants as intermediate tion on infected gasters is lacking. However, hosts, including the grass-climbing behaviour given what is known of the natural history of of Formica spp. infected with the fluke Dicro- C. atratus (especially their frequent foraging on coelium dendriticum, the yellow colour of ces- bird faeces; reviewed by de Andrade and Bar- tode-infected Leptothorax spp., and the oni-Urbani 1999) and circumstantial evidence distended gasters of fluke-infected Campono- from field experiments (Yanoviak et al.2008b), tus spp. These and many other examples are fruit mimicry remains the most parsimonious summarized in reviews by Schmid-Hempel explanation. Many Neotropical angiosperms (1998) and Moore (2002). Evolutionarily, these have small red fruits available at different times changes in host appearance or behaviour are of year, and it is logical that a bird foraging on often interpreted as extended phenotypes of such fruits would sample any similar red object the parasites (Dawkins 1982; Hughes et al. in its vicinity. 2008). A plausible alternative hypothesis to fruit Recently, a striking case of ant manipulation mimicry is that the red gasters make C. atratus by a parasite was uncovered in the rainforest workers more conspicuous to predators. Such canopies of Panama and Peru. Workers of the ‘increased conspicuousness’ strategies are arboreal ant Cephalotes atratus infected by common among parasites, although few have the nematode Myrmeconema neotropicum been studied experimentally (Moore 2002). have red gasters containing several hundred Increased conspicuousness is not supported in worm eggs (Poinar and Yanoviak 2008; Yano- this case for at least two reasons. First, C. viak et al. 2008b). The life cycle of the nema- atratus is already one of the most conspicuous tode is closely linked to the life cycle and arboreal ant species in Neotropical lowland temporal polyethism of the ant; peak redness rainforests. Aside from non-selective foraging occurs when the ant is spending large amounts by tropidurid lizards, the workers are gener- of time outside the nest, and coincides with ally ignored by insectivorous vertebrates (de the presence of infective nematode larvae Andrade and Baroni-Urbani 1999; S. Yanoviak, within the eggs. The colour change is not personal observation). Thus, although in- caused by the deposition of red pigments. fected workers stand out from healthy work- Rather, it results from localized exoskeletal ers, this difference is unlikely to greatly thinning or leaching of pigments by the de- increase predation on a common but unpal- veloping worms. This dramatic change in ap- atable ant that is already an easy prey. pearance is accompanied by continuous Second, the colour red is generally apose- gaster-flagging and a substantially weakened matic in insects. To overcome this strong neg- postpetiole, characteristics not found in ative signal, infected ants should resemble healthy ants. During the latter stages of in- non-insects, or red gasters should provide a fection, the parasitized ant becomes sluggish tasty reward. At the peak of infection (Plate 4), and assumes an erect posture (Plate 4). parasitized workers are practically immobile. In combination, these changes likely facilitate They resemble ants morphologically, but not the consumption of ant gasters by frugivorous behaviourally. Given that nematode eggs pass or omnivorous birds, which presumably mistake through birds undigested (Yanoviak et al. continues ANTS AS MUTUALISTS 99

Box 6.1 continued

2008), there is no obvious reward (nor penalty, ant–plant and ant–fungal mutualisms support excluding effort) associated with consuming an entire research programs, books, and confer- infected ant. Thus, a fundamental assumption ence symposia. In contrast, ant symbioses with of the increased conspicuousness hypothesis – nematodes (Poinar et al. 2006) are under- that an attractive signal is associated with investigated. Such parasitism is frequently valuable resources – is not supported. Likewise, overlooked or mistaken as a taxonomic variety, if there is no negative consequence of gaster as occurred with red-gastered C. atratus over a consumption (a sting or noxious chemical), this century ago (Poinar and Yanoviak 2008). mistake should persist in the bird’s behavioural ‘Berry’ ants exemplify the remarkable inter- repertoire. connectedness of species in tropical forests, Symbioses between ants and other and hopefully will stimulate additional organisms are common and well documented; research on ant–parasite interactions.

dependency in these interactions (Section 6.5), and dler and Dixon 2005). Within aphids, 40% of species put this variation in the context of macroevolution- are ant-tended, and many aphid genera include ary variation (Section 6.6). Finally, we highlight the both tended and untended species (e.g. Mooney et utility of these interactions for addressing questions al. 2008). These hemipterans tap into host plant fundamental to the field of ecology (Section 6.7) and phloem sap, which is rich in carbohydrates but conclude (Section 6.8) by identifying promising relatively poor in nutrients and amino acids. As a areas of future research. consequence, sap-feeding hemipterans must dis- pose of large quantities of processed, but nonethe- 6.2 Ants providing protection for food less sugar-rich, fluid. Many ants collect this sugary liquid waste, commonly referred to as honeydew. Trophobiotic interactions involve the consumption Ant attendance often results in larger hemipteran of a food reward, often in return for protection from colonies (Way 1963) and greater fecundity (Bristow natural enemies. For ant-loving hemipterans, cater- 1983; Del-Claro and Oliveira 2000). Ants that other- pillars, and most plants, these rewards almost in- wise prey upon arthropods do not attack the sap- variably involve a sugary and/or nutrient-rich feeding herbivores, or at least do so more rarely. liquid, one that is collected by the foragers that However, the incentives to view some proportion patrol the area surrounding the resource (Plate 3). of an aphid colony as prospective prey rather than Highly specialized ant-plants (myrmecophytes) mutualistic partners may increase as honeydew- offer additional food rewards and provide ants supplied carbohydrates become less limiting with with a domicile. colony growth (Cushman 1991; see Figure 6.1). In addition to this occasional predation, hemipterans may also bear yet unrevealed ecological or physio- 6.2.1 Sap-feeding hemipterans logical costs from their mutualisms with ants. For Many ant species engage in mutualisms with her- example, in the absence of predators, ants can re- bivorous hemipterans (Stadler and Dixon 2005; duce aphid reproduction (Stadler and Dixon 1998, Way 1963), and 41% of ant genera include tropho- Yao et al. 2000). biotic species (Oliver et al. 2008). ‘Myrmecophily’ The mechanism of ant benefits to tended hemi- (ant-loving) occurs within most families of the Ster- pterans is most often presumed to occur via protec- norrhyncha and Auchenorrhyncha (Hemiptera), in- tion from natural enemies (Buckley and Gullan cluding aphids (Aphididae), coccids and scales 1991; Stadler and Dixon 2005; Way 1963). Ants pro- (Coccoidea), and membracids (Membracidae) (Sta- vide protection against enemies that are frequently 100 ANT ECOLOGY

water, carbohydrates are the dominant constituents of hemipteran honeydew. However, sugar type, AR NE nutrients, and plant secondary compounds all in- FAR fluence the attractiveness and presumably nutritive ND MA value of honeydew for ants (Blu¨ thgen et al. 2004b; DE + IN TE see Figure 6.1 and Chapter 7). Because direct mea- RO P H IG sures of ant fitness are rare (but see Helms and H 0 Vinson 2008), we can only infer that honeydew attractiveness is indicative of nutritive value and - value to the colony. Because ants are effective and abundant preda- Behaviour of ants toward herbivores tors of many arthropods, ant–hemipteran mutual- isms have been defined as a ‘keystone interaction’ LOW Sp1 Sp2 HIGH (Styrsky and Eubanks 2007) where variation in the Nutrient content and/or quantity of herbivore attractants strength or occurrence of the interaction has far- reaching consequences for the community in An ant’s response to prospective mutualists Figure 6.1 which it is embedded (see Bishop and Bristow may vary with the prospective benefits and costs of interacting with that partner. Cushman (1991) proposed 2001; Kaplan and Eubanks 2005; O’Dowd et al. that incentives for antagonistic interactions towards ant- 2003; Wimp and Whitham 2001). With respect to tended herbivores (e.g. predation) will increase with herbivores, hemipteran-tending ants increase mu- travelling costs, the ant colonies’ demand for protein, and tualist abundance while often preying upon un- where the quality or quantity of rewards offered by the tended herbivores (e.g. Bishop and Bristow 2001; prospective mutualists is low. When reward quality is high (e.g. Sp2 relative to Sp1), protein demand is low, and Mooney 2007). Key questions for the ecology of prospective partners are close to nests, ants are more ant–hemipteran mutualisms have been whether likely to act as mutualists. (Reproduced with permission, the net effect of tending ants is to increase or de- from Cushman 1991). crease total herbivore abundance, and to what indi- rect effect on plant growth and fitness (e.g. Horvitz somewhat specialized as hemipteran predators, in- and Schemske 1984). Thus far, the literature sug- cluding ladybird beetle larvae and adults (Coleop- gests that hemipteran-tending ants typically de- tera: Coccinellidae), syrphid fly larvae (Diptera: press the local abundance and species richness of Syrphidae), lacewing larvae (Neuroptera: Chryso- several guilds of chewing herbivores, often to the pidae), and parasitoid wasps (Hymenoptera: Bra- host plant’s benefit (Styrsky and Eubanks 2007). conidae), although ants can also provide protection There are, however, spectacular exceptions to this against more generalist predators such as spiders generalization (e.g. Box 15.1; O’Dowd et al. 2003). (Cushman and Whitham 1989; Del-Claro and Oli- veira 2000). Other benefits can include reduced 6.2.2 Lepidopterans fouling from honeydew accumulation (Bach 1991), reduced competition from other herbivorous in- Approximately 70% of Lepidoptera in the family sects (Smith et al. 2008), and allowing aphids to Lycaenidae (an estimated 6,000 species) whose life divert resources away from predator avoidance or histories are known engage in associations with ants parental care, and towards feeding, growth, and (Eastwood et al. 2006; Fiedler 2006; Pierce et al. reproduction (Abbot et al. 2008; Bristow 1983; Flatt 2002). Although some species of lycaenid may act and Weisser 2000). Such non-protective benefits are as parasites of ants, we focus on the mutualistic rarely studied, and their frequency or importance species (Travassos and Pierce 2000). The lycaenid compared to protection from predators is not well larvae and/or pupae attract the attention of ants by understood. producing nitrogen-rich secretions (e.g. Agrawal It would seem that the entire benefit of tending and Fordyce 2000; Devries 1991), in some cases hemipterans for ants is nutritional. Aside from complemented by chemical and acoustic signalling ANTS AS MUTUALISTS 101

(Devries 1991; Travassos and Pierce 2000). Perhaps a as a result of the nitrogen investment in reward production, species that feed on nitrogen-fixing plants or on nitrogen-rich plant parts such as flow- ers and seed pods are more likely to be ant-tended than are species that feed on other plant types or parts (Pierce 1986 but see Billick et al. 2005; Fiedler 1995). As with ant-tended hemipterans, ants benefit lycaenids by reducing attacks by parasitoids and predators (e.g. Devries 1991; Pierce and Mead 1981; Wagner and Kurina 2003). Lycaenids may add mass and reach maturity more quickly when protected by ants (Cushman et al. 1994), and some species preferentially oviposit in sites where ant b densities are high (Wagner and Kurina 2003).

6.2.3 Extrafloral nectary-bearing plants

Plant species in over 90 families attract ants to nec- taries not associated with flowers (hereafter, extra- floral nectaries or EFNs; Koptur 1992). These structures typically produce carbohydrate-rich nec- tar that can also include trace amounts of nitrogen or amino acids (Koptur 1992; Chapter 7). The nec- tar-attracted foragers may subsequently patrol the plant in search of further nectar, to deter competi- Myrmecophytes offer domatia in a variety of tors, and/or to consume alternative food items. Figure 6.2 forms, many requiring that the ants gain access to a This combination of foraging, deterrence, and con- hollow structure by cutting through plant tissue. (a) A sumption can decrease or alter the distribution of Pseudomyrmex spinicola worker at the entrance of its herbivory (Chamberlain and Holland 2009; Heil et nest on a swollen thorn Acacia. The thorn is hollow but al. 2001; Koptur 1992; Ness 2003a; Oliveira et al. the ants must cut a hole to gain initial entrance. (b) An queen cutting into the soft tissue of a 1999; Rudgers and Strauss 2004; and see appendices Azteca isthmica Cecropia tree to start a nest in the hollow center. (Photos: in Rico-Gray and Oliveira 2007), typically to the net Alex Wild) benefit of most ant-tended plant partners (Cham- berlain and Holland 2009). The interaction between EFN-bearing plants and gen et al. 2004b; Dı´az-Castelazo et al. 2004; Oliveira ants often includes multiple ant species sympatri- et al. 1999). cally foraging on the same plants or plant species in the same population (e.g. Cuatle et al. 2005; Oliveira 6.2.4 Myrmecophytes (ant-plants) et al. 1999; Schemske 1980). These foraging bouts may be segregated in space or time (i.e. within Plants in over 100 tropical genera host ants in particular branches, diurnal versus noctural, within specialized structures such as swollen thorns, hol- particular temperature ranges or seasons; Dı´az- low stems, and leaf pouches, collectively termed Castelazo et al. 2004; Oliveira et al. 1999; Rico-Gray domatia (Bronstein et al. 2006; see Figure 6.2). 1993; Schemske 1980). As a result, the EFNs on a These myrmecophytes or ‘ant-housing plants’ can particular plant may provide an important resource also provision ants with lipid and protein-rich food for an ant community even if it is not particularly bodies and/or nectar (Heil and McKey 2003; important for any one population or colony (Blu¨ th- O’Dowd 1982), or host honeydew-producing 102 ANT ECOLOGY

a myrmecophytic Hirtella spp. and Tachigali myrmeco- phila do not produce food rewards but instead sup- port hemipterans that nourish the resident ant colonies (Rico-Gray and Oliveira 2007). Myrmeco- phytic interactions are much less widespread than facultative ant–plant associations involving EFNs (Heil and McKey 2003) and exhibit much greater specialization by both partners. For example, some myrmecophytic Acacia offer extrafloral nectar high in invertase (sucrose-cleaving enzyme) and low in sucrose, which corresponds to the preference of resident Pseudomyrmex ants for su- crose-free nectar that is unpalatable to other ants (Heil et al. 2005). Morphological adaptations in- b clude the prostoma (unlignified organ at the tip of the domatia) of Leonardoxa plants, the shape, and size of which corresponds strongly to the head of mutualistic ants (Brouat et al. 2001) and the wax crystals on the stems of some Macaranga that ex- clude ants not adapted to the slippery surface (Fed- erle et al. 1997). Plant-dwelling ants may provide nutrients, and/ or protect their hosts from invertebrate and verte- brate herbivores, plant pathogens, and encroach- ment by competing plants (Bronstein et al. 2006; Davidson and McKey 1993; Heil and McKey 2003). These benefits can be pronounced. For exam- Figure 6.3 Myrmecophytes differ in the types of food ple, a successful ant–plant symbiosis can create they offer to their resident ants. (a) Lipid- and glycogen- large monospecific ‘devil’s gardens’ within other- rich Mullerian food bodies on a Cecropia tree. (b) A wise diverse tropical rainforests (Frederickson et al. Pseudomyrmex spinicola worker on a swollen thorn Acacia harvesting a protein-rich food body to feed to the 2005). Plants that house ant residents may also (or colony’s larvae. (Photos: Alex Wild) instead) benefit from greater access to nitrogen and

CO2 as a result of the activities of its plant-dwelling ants (Sagers et al. 2000; Treseder et al. 1995). The hemipterans (Gaume et al. 1998; Palmer et al. 2008) benefits of its resource transfers may exceed the (see Figure 6.3). Some of the best-studied myrme- value of any protection provided by the ants in cophytic relationships include those of Acacia with some systems and/or ecological settings. Pseudomyrmex ants (e.g. Janzen 1966, 1967a), Cecro- pia with Azteca spp. and other ants (e.g. Folgarait 6.2.5 The best ant partners and Davidson 1994, 1995; Longino 1989), and Piper with Pheidole ants (Letourneau 1983; Letourneau et From the trophobiont’s perspective, the ideal pro- al. 2004) in the Neotropics, and Macaranga with tectors are competitively dominant ants capable of Crematogaster ants in southeast Asia (e.g. Feldhaar aggressive behaviours (biting and stinging) and et al. 2003; Fiala et al. 1989; Itino et al. 2001). Myrme- mass recruitment (e.g. Buckley and Gullan 1991) cophytic Acacia, Macaranga, Cecropia, and Piper that might deter the partner’s natural enemies. Ant provide their ant partners with domatia and characteristics that lessen the costs of foraging or food bodies (see reviews in Davidson and McKey patrolling can also increase the likelihood of mutu- 1993; Rico-Gray and Oliveira 2007), whereas alistic interactions by allowing a trophobiont that ANTS AS MUTUALISTS 103 produces modest rewards to, nonetheless, engage Cost ants in a beneficial manner (see Figure 6.1). For Maximum net example, foraging costs are reduced if the prospec- Benefit tive ant partners are capable of establishing satellite benefit nests at the base of plants with EFNs or hemipteran aggregations, or if the plant itself is the domicile of the colony. Because trophobiosis invariably in- Cumulative energy volves the collection of sugary and/or nutrient- (mg[carbon] / plant) rich liquid, key adaptations include the capacity for trophallactic exchange of liquid food among Mutualism Parasitism members of the colony (Fiedler 2006) and morpho- N Max logical changes to the ant’s proventriculus and gas- Colony size ter, which enable them to carry large amounts of Figure 6.4 Graphic model of the cost and benefit to a sugary fluids (honeydew and nectar) and regulate myrmecochore of hosting an ant colony of varying size. the digestion of these fluids (Davidson et al. 2004). The model describes a system wherein the cost of This is one explanation why subfamilies such as maintaining and housing ants increases linearly with colony size, while the benefits that ants provide saturate. Formicinae and Dolichoderinae (and, more rarely, N is the colony size that provides the greatest difference Myrmicinae), whose members have some subset of between benefit and cost (i.e., greatest net benefit) to the these adaptations, are the most common tenders, plant. Beyond a threshold ant colony size (Max), plant and why these foragers may tend to more than one costs exceed plant benefit. Ant colonies at this stage may of these partners within a particular habitat (e.g. experience disproportional net benefit and/or be subject to plant reprisals. (Reproduced with permission, from Blu¨ thgen et al. 2000; Devries 1991; DeVries and Fonseca 1993). Baker 1989; Rico-Gray 1993). From the perspective of the myrmecophyte, the ideal ant partner is quick to detect and deter would- incremental benefit to a plant of hosting additional be plant antagonists at a minimal cost to the plant. ants progressively lessens as ant density increases However, because the ecologies of myrmecophyte and the structural or metabolic costs of hosting and plant-ant are largely inseparable, among-sys- those ants increases linearly (i.e. each ant costs the tem transplants that could allow scientists to con- same), the net benefit of hosting ants could lessen as trast the benefits of particular pairings are colonies increase in size (Figure 6.4). This ant–plant impossible. That is, we cannot test whether Acacia conflict can influence the density of plant-ants that plants might do better hosting the Azteca ant associ- can occupy a given plant, population or community ates of Cecropia. However, the diverging natural (Fonseca 1999), and may well dictate the best part- histories of particular systems provide clues to ex- ner for particular settings. plain why the favoured (or at least realized) char- The worst partner ants decrease the fitness of acteristics may differ among systems. Fast-growing their partners. For example, in some settings, ants pioneer trees with rapid rates of resource supply, may consume more aphids than they protect from such as Acacia, Macaranga, and Cecropia, often host natural enemies (see Figure 6.1). Other costs might an active, aggressive workforce of large ants (Da- be more subtle. Highly aggressive ants that visit vidson and McKey 1993). Smaller trees and shrubs, EFNs may also deter pollinators as effectively as such as Leonardoxa and Piper, often host smaller, they do to natural enemies (e.g. Ness 2006), leading more timid or sluggish workers that can nonethe- to conflict between the defensive and reproductive less be effective against very small herbivores, eggs, mutualisms. Some ant residents also prune the and microbes (Gaume et al. 1997; Letourneau 1983). flowers of their myrmecophytic hosts (Stanton Food (Itino et al. 2001) and nesting site (Fonseca et al. 1999; Yu and Pierce 1998) in an apparent effort 1993) resources impose limits to the hosting capaci- to reallocate host resources towards ant rewards, ty of plants and imply a trade-off between the num- enabling increased colony size at the plant’s ex- ber and size of ants that can be hosted. If the pense (see Figure 6.4). Ant-tended plants may 104 ANT ECOLOGY

et al. 1994). As a result of the nutritive value and chemical signalling component of the elaiosome and a durable seed coat, ants that might otherwise act as seed consumers are perhaps converted into elaiosome consumers, and hence, seed dispersers (Rico-Gray and Oliveira 2007). Interestingly, plants may also co-opt the attention of carnivorous ants; the most avid collectors of elaiosome-bearing seeds rarely include plant material in other aspects of their diet (Hughes et al. 1994). Myrmecochorous species are found in >80 plant families, and the Figure 6.5 Ants from the genus Rhytidoponera are morphological features associated with myrme- important seed dispersers. Here, Rhytidoponera metallica cochory have evolved at least 20 times in the mono- carries a seed with elaiosome attached. (Photo: Benoit cots (Dunn et al. 2007a). This estimate may be Gue´ nard) conservative, as some seeds that rely on ants for dispersal lack food rewards. For example, the dia- spores of some ‘ant garden’ plants use odorants, limit these indirect costs by including ant-deterring rather than food rewards, as ant attractants (e.g. compounds in their flowers (e.g. Ness 2006; Will- Youngsteadt et al. 2008). Whether that collection mer and Stone 1997a). provides sufficient benefit to the ants to qualify as a mutualism is unclear. The conventional forms of myrmecochory benefit 6.3 Ants provide dispersal for food the ant colony by providing a food resource that can 6.3.1 Seeds enhance the colony’s reproductive output (Gam- mans et al. 2005; Morales and Heithaus 1998). Myr- Myrmecochory is the dispersal of ant-adapted seeds mecochores produce their seeds in seasons where by ants. Over 90% of the >3,000 ant-dispersed plant seed collection by ants is most likely to occur. This species are found in the South African fynbos and is the early summer in temperate deciduous forests, in areas of Australia dominated by sclerophyllous when ant forgers are both highly active and have plants (Berg 1975; Bond and Slingsby 1983). Most of dietary preferences that make elaiosomes attractive the remaining identified ant-dispersed species are (Oberrath and Bohning-Gaese 2002), and plants spring ephemerals in the temperate deciduous for- have few opportunities for interactions with avian ests of northern Europe, Japan, and North America; frugivores (Thompson 1981). The benefits to the myrmecochores account for 40% of the herbaceous seed include protection from granivores (e.g. Bond species and 60% of emergent stems in portions of and Slingsby 1984; Christian 2001; Turnbull and temperate deciduous forests of the eastern United Culver 1983) and/or fire (Christian 2001), directed States (Beattie and Culver 1981; Handel 1981). dispersal to atypical microsites (such as nutrient- Myrmecochorous seeds have an attached, lipid- rich ant middens; Davidson and Morton 1981, but rich food reward, called an elaiosome, which at- see Rice and Westoby 1986), dispersal away from tracts ant foragers (Figure 6.5). Because the elaio- parent plants and siblings (e.g. Bond and Slingsby some’s fatty acid composition is similar to that of 1984; Horvitz and Schemske 1986; Kalisz et al. 1999; insect prey (Hughes et al. 1994), the diaspore (seed Ness et al. 2004), and enhanced germination rates þ reward) is attractive to omnivorous foragers. (e.g. Cuatle et al. 2005). The cumulative effect can be Ants may preferentially collect seeds with larger greater fitness for those seeds that are collected by elaiosomes or more favorable elaiosome-to-seed ants (Hanzawa et al. 1988). Although these conse- ratios (Mark and Oleson 1996), and some elaio- quences are often studied in isolation, multiple ben- somes also include compounds that elicit collection efits may be derived from any one ant–seed behaviors by workers (e.g. 1,2-diolein in Hughes interaction (Giladi 2006). ANTS AS MUTUALISTS 105

6.3.2 The best ant partners between plants and legitimate pollinators. These consequences are the likely selection pressures for Changes in the seed-dispersing ant communities the chemical and physical impediments that can can alter seed survival rates, mean and maximum deter ants from entering flowers (e.g. Galen and seed dispersal distances, and the distribution and Butchart 2003; Ness 2006). composition of mature plant communities (Ander- Plant characteristics that can favour pollination sen and Morrison 1998; Bond and Slingsby 1984; by ants (or increase the incentives for ant pollina- Christian 2001; Ness et al. 2004; Ness and Morin tion) include living in sites where ant activity is 2008; Parr et al. 2007; Chapters 8 and 15). From the high (and/or other pollinators are rare), few syn- perspective of a myrmecochorous plant, ideal ant chronously blooming flowers per plant (to mini- partners share several characteristics. High quality mize intra-plant pollination or stigma-clogging for dispersers are typically solitary, omnivorous fora- self-incompatible plants), pollen volumes insuffi- gers that range far from their nest, disperse dia- cient to elicit grooming behaviours by the ants, spores at substantial distances to those nests (the and nectar rewards sufficiently unrewarding to dis- criteria for ‘substantial’ may be defined by the size courage visitation by alternative, more expensive, of plants and the scale of soil heterogeneity within pollinators (Hickman 1974). Although rare, such the site), feed on the elaiosome while leaving the systems do exist. There are also a few plant species seed intact, and bury the seeds shallowly in micro- that receive pollination services by mimicking op- sites where they can respond to germination cues portunities for ant copulation (e.g. Leporella fimbriata (Giladi 2006). Low quality dispersers may be gra- orchids are pollinated by male Myrmecia urens, Pea- nivorous, disperse seeds at insufficient distances to kall 1989). avoid competition with maternal and sibling plants, dissect diaspores in situ (i.e. ‘elaiosome rob- bing’) rather than carry them to the nest, and cache 6.4 Ants, fungi, and bacteria high densities of seeds deep underground where Originating 50 Mya (Schultz and Brady 2008), the germination is unlikely (Giladi 2006). Well-studied tripartite association among ants, fungal cultivars, ‘high quality’ ants that collect a disproportionate and actinomycete bacteria is perhaps the most high- amount of myrmecochorous seeds are Rhytidopo- ly evolved and complex set of mutualisms in ant nera spp. in Australia (Figure 6.5; Andersen and ecology. More than 210 species in 13 genera of Morrison 1998; Gove et al. 2007; Hughes et al. Myrmicine ants in the New World Attini tribe cul- 1994) and the Aphaenogaster rudis complex in tivate basidiomycete fungi as their main food North America (Beattie and Culver 1981; Ness and source by collecting and preparing an appropriate Morin 2008). fungiculture substrate (Currie 2001; Poulsen and Currie 2006). The ant genera vary in their choice of fungiculture substrate, colony size, and polymor- 6.3.3 Pollen phism. The more basal or ‘lower’ attines utilize The ubiquity of ants and their diverse interactions insect corpses, faeces, or plant detritus as fungal- with plants begs the question of why ants so rarely growing substrates, and tend towards smaller, act as pollinators. Several characteristics make ants monomorphic colonies. In contrast, the more poor candidates: maximum foraging distances are derived or ‘higher’ attines utilize plant detritus or short relative to winged visitors, ant territoriality fresh plant material (e.g. leaf-cutting ants; Figure may decrease the likelihood of outcrossing among 6.6), can display extreme polymorphism, and may plants, and exposure to ants can reduce the viability achieve colony sizes of several million individuals of pollen (likely due to ant-borne antibiotics; Beattie (Currie 2001; Poulsen and Currie 2006). Queens and et al. 1984). In so far as these shortcomings decrease larvae of attine ant colonies feed exclusively on the the success of both male and female plant function fungus, while workers may supplement their fun- (e.g. Galen and Butchart 2003), flower-visiting ants gal diet with plant sap (Quinlan and Cherrett 1979). may be unwelcome ‘parasites’ of the interaction In the case of leaf-cutting ants, the fungi convert 106 ANT ECOLOGY

a Ants + + – –

+ + + Bacteria Cultivar – – –

+ – + Black yeast Parasite +

Figure 6.7 A diagram of the direct and indirect interactions of the attine ant-microbe symbiosis. Solid b lines represent direct effects, dashed lines represent indirect effects, requiring the presence of an intermediary species. Cost (-) or benefit (+) deriving from the interaction is indicated at the tip of the arrowhead. Cultivar = fungal cultivar; parasite = specialised fungal parasite, Escovopsis; bacteria= actinomycete bacteria, Pseudonocardia, hosted on the ants; and black yeast= parasite of the bacteria. (Modified with permission, from Little and Currie 2008).

duced proteolytic enzymes around the fungal gar- den, and maintaining the garden chamber at the appropriate temperature and humidity (Poulsen and Currie 2006). The ants employ behavioural and are among the most conspicuous ants in Figure 6.6 Atta chemical means to protect their fungal gardens from the Neotropics and their colonies can number millions of other microbes. The use of a platform by founding workers in multiple subcastes. (a) An Atta cephalotes worker carries its harvest back to the nest to feed the Atta queens reduces the risk of infection by microbes colony’s fungal cultivar. (b) Atta cephalotes workers tend in the soil (Ferna´ndez-Marı´n et al. 2007). Weeding the colony’s fungal garden. (Photos: Alex Wild) and grooming by workers also reduce contamination by non-mutualist microbes (Currie and Stuart 1991). inedible plant material into lipid and carbohydrate- Metapleural gland secretions provide effective gen- rich gonglydia, making the monophagous ants eral antibiotics and defend the fungal cultivars from ‘ecologically polyphagous’ (Rico-Gray and Oliveira an array of microbes (Poulsen et al. 2002; see Box 9.1). 2007). As a result, the ant–fungal composite feeds The weeding and grooming behaviours and me- on a great diversity of widely distributed plants tapleural gland secretions are not effective against that would otherwise be inaccessible to the fungi specialized fungal parasites in the genus Escovopsis. and/or inedible to the ants. Thus the ants and their fungi depend on another Fungus-cultivating ants have an elaborate set of mutualist, actinomycete bacteria (Figure 6.7). These behaviours and traits that facilitate fungal cultiva- actinomycetes, in the genus Pseuodonocardia, are tion. The selection of an appropriate substrate is key reared in specialized, elaborate crypts present in to fungal growth. Leaf-cutting ants avoid harvesting genus-specific locations on the cuticles of attines from plants with incompatible chemistry, possibly (Currie et al. 1999, 2006). The bacteria produce anti- via feedback from the fungus (North et al. 1997). biotics that selectively inhibit the growth of Escov- Attines further promote the growth of their fungal opsis (Currie et al. 1999) and are associated with all cultivars by pruning, redistributing fungus-pro- attine ants that have been examined (Currie et al. ANTS AS MUTUALISTS 107

2006). In Acromyrmex, the bacteria are most abun- populations of the participants. That is, might the dant on major workers that are most active at the positive feedbacks derived from these interactions bottom of the fungal garden where Escovopsis is encourage these populations to grow progressively most likely to be encountered (Poulsen et al. 2002). larger ad infinitum? At least three explanations clari- The actinomycetes further benefit the fungal culti- fy why this ‘orgy of mutual benefaction’ is so rarely vars by providing growth-promoting compounds observed (but see ‘invasional meltdown’ as in (Currie et al. 1999) and may also directly benefit the O’Dowd et al. 2003 and Box 15.1). First, other forces ants by protecting them from pathogens (Currie external to the mutualism, including natural enemies 2001). Benefits conferred on Pseudonocardia by the attracted by the success of one participant, intraspe- ants include dispersal (by virgin queens during the cific competition, or abiotic factors, may eventually nuptial flight), provision of a unique habitat in the limit the populations of at least one partner. For cuticular crypts, and nourishment from specialized example, the black yeast symbionts that exploit the glands (Currie 2001; Currie et al. 2006). However, mutualism between fungus-growing ants and their symbiotic black yeasts can parasitize Pseudonocar- actinomycete bacteria decrease the ability of the ants dia, reducing its growth and decreasing the ability to protect their fungal gardens from the Escovopsis of the ants to suppress Escovopsis infection of their parasite (Little and Currie 2008), to the detriment of fungal gardens (Figure 6.7; Little and Currie 2008). fungal garden health (Currie 2001). Second, the re- The synergism between leaf-cutting ants and sources on which the interactions are based may leaf-digesting fungi provides perhaps the most dra- collapse, as could be the case if a synergistic ant– matic example of the community-wide conse- aphid interaction was overly detrimental to the host quence of ant mutualisms. This ant-fungal plant. Third, the benefits conferred by participating collaboration exploits up to 50% of the plant species in the interaction may saturate. For example, from (Cherrett 1968, 1972) and can remove >10% of total the perspective of a plant or honeydew-producing leaf production in some Neotropical forests (Ho¨ll- aphid aggregation, the distinction between being dobler and Wilson 1990). Few other herbivorous tended by 5 ants versus 10 may be of negligible animals approach this breadth or magnitude of importance if 3 ants are sufficient to provide services impact. Not surprisingly, this consumption can required to increase their population (Ness et al. profoundly constrain plant recruitment and direct 2006). Additional ants may even be worse, if the the nature and pace of plant succession (Vasconce- costs of reward production increase linearly with los and Cherrett 1997; Wirth et al. 2003), and the partner abundance (Fonseca 1993; see Figure 6.4). consolidation of resources in and around leafcutter Likewise, the carbohydrate rewards provided to nests can alter soil properties and the distribution of these ants may become progressively less important in-soil resources (Farji-Brener and Illes 2000; Wirth to the well-being of the colony once access to protein, et al. 2003). rather than carbohydrates, becomes limiting (see Fig- Fungiculture in non-attine ants is much less stud- ure 6.1). ied. Ants in the Old World Lasius genus, in the Case studies of ants and their ‘trophobionts’ have subgenera Dendrolasius and Chthonolasius, utilize shown remarkably disparate costs and benefits ascomycete fungi to bind shredded wood or soil when studied in different settings. The benefits to reinforce nest walls. The ants nourish their provided by ants to aphids have been shown to fungi with honeydew and protect them from com- decline with aphid colony size (Breton and Ad- peting fungi, possibly through grazing (Schlick- dicott 1992). Because aphids can bear costs for pro- Steiner et al. 2008). visioning mutualist ants (Stadler and Dixon 1998, Yao et al. 2000), the net benefit of ants can be pre- 6.5 Context dependency and stability dicted to depend on the risks from natural enemies, honeydew fouling, and competition from other her- Because mutualisms are, by definition, reciprocally bivores. Moreover, aphids can compete intra- and beneficial interactions, it is reasonable to ask what interspecifically for ant attendance, and whether forces stabilize these interactions and regulate the ants are mutualists for a particular aphid clone 108 ANT ECOLOGY depends on the availability of other sources of hon- promote additional ant attendance include greater eydew and nectar (Cushman and Addicott 1989). production of extrafloral nectar (Heil et al. 2001; There is also evidence that host plants can mediate Ness 2003a), ant domiciles (Stanton et al. 1999), not only the strength of ant benefits to aphids, but lycaenid food rewards (Agrawal and Fordyce also the direction of ant effects (Mooney and Agra- 2000), or acoustical (Morales et al. 2008a) and chem- wal 2008). Variation in phloem sap quantity or ical (Del-Claro and Oliveira 1996) signalling to ants quality may be responsible for mediating these during times of need. ant–aphid interactions (see also Figure 6.1). Some level of constancy in partner identity, Context-dependency might be particularly prev- quality, and benefit is essential if local evolution- alent in interactions where ants primarily provide ary specialization for particular mutualisms is to benefit by conferring protection. Most ant visitors occur at the population level. Some myrmeco- (or at least visits) may not benefit the myrmecophile philes will occur in settings where interactions because: (a) the workers do not protect the partner with ants are more necessary, and hence, more (i.e. when ants are timid or ineffectual against ant- beneficial (Rudgers and Strauss 2004). Among- adapted herbivores) or (b) the partner’s need for site variation that is stable through time could protection does not coincide with interactions with result in local evolution if the effective popula- particular ant species or populations (e.g. EFNs: tion size of partners is small relative to the area Schemske 1980; hemipterans: Cushman and Whi- occupied by a particular partner taxon (Horvitz tham 1989; myrmecochores: Fedriani et al. 2004). and Schemske 1990; Rudgers and Strauss 2004), Perhaps as a result of selection pressures to increase whereas temporal variation can only lead to dif- the likelihood that ants can provide appropriate fuse selection by the assemblage of mutualists services when that service is required, the spatio- (‘the interaction’) rather than particular partner temporal distribution of reward production is cor- species. There is some evidence that the identity related with the plant’s vulnerability to natural of ants that act as prospective mutualists varies enemies. For example, EFNs are often located in greatly among sites but can be more consistent areas where the consequences of herbivory could over time within sites than are other mutualisms be severe (e.g. young leaves, at the base of repro- (e.g. see EFN tending ants versus pollinators in ductive units; Horvitz and Schemske 1990; Oliveira Horvitz and Schemske 1990). This may be due to et al. 1999; Schemske 1980). Likewise, myrmeco- the longevity (and immobility) of individual co- chores drop seeds during the day, when foraging lonies, relative to some of their partners. by granivorous rodents is lessened and the likeli- hood of seeds being collected by ants is greatest 6.6 Macroevolutionary patterns in the (Cuatle et al. 2005; Turnbull and Culver 1983). face of variation If mutualist ants, or the subset of ants that are particularly effective, are a limited resource, pro- Although over half the ant subfamilies do not in- spective partners will compete for their services clude species known to engage in mutualisms with and a subset may suffer from decreased service. trophobiotic insects, mature plants (EFN-bearing or For example, experimental augmentations of mem- myrmecophytic) or fungi, the incidence of all three bracid aggregations decreased overall tending rates interactions are positively correlated with one an- by ants due to the decrease in the ratio of ants to other among the remaining, vigorously mutualistic, membracid. The consequence of this decrease in subfamilies (Oliver et al. 2008). Within those subfa- service was a >90% decrease in the production of milies, however, mutualisms with trophobiotic in- membracid adults (Cushman and Whitham 1991). sect or plants are negatively correlated with those A shortage of mutualists, or the disincentives of with fungi at the genus level. One explanation is supporting partners when they are unnecessary, that ant lineages need to specialize in one type may explain why some myrmecophillic partners of mutualism when the adaptations for service or have adaptations to help them attract additional receiving benefit in one mutualism diverge from ant partners. Some of the methods employed to another. Specifically, the characters that favour ANTS AS MUTUALISTS 109 collecting fungal substrates and maintaining suit- The strongest evidence of coevolution and ‘part- able fungal growing conditions in return for edible ner-filtering’ occurs in the interactions among mycelia differ from those of aggressive defense and myrmecophytes and their partners (see 6.2.4, Bron- ingesting sugary secretions (Oliver et al. 2008). stein et al. 2006; Brouat et al. 2001; Federle et al. 1997; Among facultative associations, there is evidence Heil et al. 2005; Janzen 1966) and between fungus- that the adaptations favouring myrmecophily are farming ants and their symbionts. All of the studied evolutionarily labile (i.e. can be acquired and/or fungus-growing ants have phylogenetically specific lost at the species level of resolution). Hemipteran modified exoskeletons for housing and feeding, for traits associated with ant tending include modifica- example, Pseudonocardia bacteria; closely related ant tion of honeydew chemical composition, aggre- species lack these modifications (Currie et al. 2006 gated feeding, longer proboscis length (Bristow but see Kost et al. 2007). That all fungus-growing ants 1991, Shingleton et al. 2005), and loss of defensive host a strain of Pseudonocardia (Currie et al. 2006), structures and predator avoidance behaviours (Sta- suggests that there is a yet-to-be-discovered mecha- dler and Dixon 2005). A complete understanding of nism for preventing establishment by other bacteria hemipteran adaptations to ant-tending is still forth- that may not act as a mutualist to the ant or the coming; some of the observed associations between fungal cultivar (Kost et al. 2007). Similarly, although traits and ant-tending are known from single the ant–fungal cultivar relationship is now thought hemipteran lineages, while associations among to be indicative of more diffuse, rather than pairwise, taxonomically disparate species do not distinguish coevolution (Mikheyev et al. 2006), the incompatibil- between evolutionary convergence (as is presumed) ity of alien fungal strains and hostile ant behaviour and common ancestry. Nevertheless, that myrme- towards alien fungal fragments can prevent the in- cophily is not constrained to any single lineage troduction of competing fungal clones (Poulsen and suggests multiple origins and high lability for mu- Boomsma 2005). tualism with ants. Consequently, many untended hemipteran species may be only a few evolutionary 6.7 Model interactions for ecology or ecological steps away from such mutualisms, and adaptations may be subtle. Among ants, adap- Ant mutualisms have several characteristics that tations that are correlated with, and perhaps favour, make them ‘model systems’ for addressing ques- trophobiosis include a modified proventriculus, tions regarding mutualism and plant defense. We polygyny, and polydomy (Oliver et al. 2008). highlight these advantages later, and propose Ant–myrmecochore interactions were not includ- promising research questions in Section 6.8. ed in the aforementioned phylogenetic analyses. In First, ant attendance and behaviour can be moni- so far as elaiosomes are dead insect analogues, their tored in real time. Ants that forage on the surface of collection and utilization by ants may require little plants, leaf litter, and soil can be counted. As a specialization or trade-offs with other mutualisms. result, variation in the number or behaviours of Further, the repeated independent origins of myr- foragers allocated to a particular task and turnover mecochory (Dunn et al. 2007a) and diversity of elaio- in the species performing a task can be accurately some shapes, weights, histological origins, caloric described. These measures can provide information and nutritional content, and manner of diaspore about the costs and benefits received by each par- presentation in that guild imply great generalization ticipant in the prospective mutualism, and how by the plants. Nonetheless, the existing field obser- these vary over time, space, or in response to exper- vations demonstrate that two ant genera collect a imental treatments. majority of the myrmecochorous seeds in sclero- Second, ants can be excluded from particular mi- phyllous Australia (Rhytidoponera spp., Figure 6.5; crosites. Many studies of ant protection mutualisms Gove et al. 2007) and temperate North America use sticky substances (e.g. TanglefootTM)toexperi- (Aphaenogaster rudis complex). Whether this consti- mentally exclude ants from some subset of their part- tutes ‘specialization’ by the myrmecochorous guild ners, or from portions of particular partners (e.g. (to say nothing of coevolution) is unclear. control versus treatment branches). Remarkably, 110 ANT ECOLOGY these exclusion treatments can even be performed at 2001; Palmer et al. 2008). Nonetheless, how often the scale of hectares (e.g. poison baits in Abbott and interactions with suboptimal partners constrain se- Green 2007). Ant densities can also be depressed by lection for the mutualism is unclear. adding ant predators (e.g. Letourneau et al.2004). Can the inclusion of suboptimal partners be benefi- Third, many individual partners interact with one cial, and are there settings in which the diversity of a ant colony for sustained periods of time. This largely partner assemblage itself confers benefits? For myr- occurs as a result of a combination of the central-place mecochores, a more diverse disperser assemblage foraging requirements of a (largely) immobile ant could increase the variety of sites where seeds are colony and territorial interactions between colonies deposited and, perhaps as a result, decrease the influ- and/or species. Although there are many exceptions ence of detrimental density-dependent processes. The to this gross generalization and the very definition of synergistic effect of multiple predators is well docu- ‘sustained’ will depend on the lifespan of the partner, mented in other systems (Cardinale et al. 2003; Sih some ant mutualisms are believed to have continu- et al.1998);thatitoccursinantprotectionmutualisms ously functioned for centuries (Frederickson et al. is a reasonable (Beattie 1985; Rico-Gray and Oliveira 2005). Irrespective of duration, this dynamic of sus- 2007), albeit largely untested, hypothesis. Further, re- tained interactions between two individuals, or at peated interactions with suboptimal partners, or in- least one individual and one superorganism, is more teractions with many of those partners, can remedy common in ant mutualisms than in those involving the mediocrity that may be so pronounced on a per more mobile partners, such as pollination. capita or per interaction basis (Ness et al. 2006). If These three attributes of ant mutualisms have fa- partner diversity does confer benefits, the costs of cilitated much of our understanding of ant and non- participating in a mutualism that typically includes ant mutualisms (Bronstein 1998; Heil and McKey a diverse assemblage may only become apparent 2003), as well as plant defense, food web structure, when a prospective mutualist is simultaneously de- and the dynamics of symbiotic interactions. For ex- prived of those diverse partner assemblages and ample, an ecologist’s ability to reliably count and limited to interacting with one or a few partners that exclude ants (or ant-occupied thorns) from particular are mediocre (or outright parasitic) in all settings. branches makes it possible to quantify and manipu- That combination of homogeneity and inadequacy late plant defenses to an extent that is nearly impos- may be a historically rare phenomenon in the natural sible (or at least terribly expensive) for chemical plant world. Or, put differently, myrmecophillic organisms defenses such as tannins or alkaloids. mayberareinhabitatswheresuchconditionsarethe norm in the natural world. We predict such pairings 6.8 Future directions may occur increasingly frequently due to anthropo- genically induced disruptions in ant faunas in the face In the following text, we highlight several ecologi- of landscape conversion, global climate change, and cal topics that we perceive as particularly critical exotic ant invasions. and promising for better understanding the role of Partner diversity and specificity of the ant-fungi- ants as mutualists. bacteria mutualism are also ongoing subjects of investigation. Recent discoveries of filamentous ac- tinomycete bacteria on non-attine ants that also 6.8.1 Diverse partners inhibit Escovopsis growth have called into question What are the consequences of interacting with a the specificity of the attine ant–actinomycete mutu- variety of prospective ant partners? There are alism and whether ants have any mechanism to many examples of great variation in partner quality control bacteria on their exoskeletons (Kost et al. (e.g. Buckley and Gullan 1991; Horvitz and 2007). In addition, much more is to be learned Schemske 1986; Miller 2007b; Ness et al. 2004; Ness about the non-attine ants that cultivate fungi for et al. 2006), and striking examples where a greater architectural purposes. Is there a similar complex frequency of interactions with suboptimal partners interplay of mutualists and parasites as has become lessens the benefits to the ant’s partner (Christian evident in the attine ant fungal gardens? ANTS AS MUTUALISTS 111

6.8.2 Benefits to ants that engage in aphid nymphs from predators), and they often lit- mutualisms erally transport those partners into their ‘sphere of influence’ (e.g. carrying seeds and aphids closer to The vast majority of studies that explore the inter- the nest). In contrast, the benefits to the ants often actions between prospective mutualists have fo- are limited to the augmentation of resources that cused almost exclusively on the consequences of are already found in the regular diet of the ant. the interaction for the non-ant partner. This dis- crepancy may be partly attributable to (a) the logis- 6.8.3 Costs and cheating tical difficulties of measuring ant fitness (but see Cushman et al. 1994; Lach et al. 2009; Morales and To answer whether the net effect of an interaction is Heithaus 1998); (b) the assumption that the partici- beneficial, our interpretation of the spatio-temporal pation of the more mobile participant (ant) is evi- heterogeneity in benefits should be balanced by an dence of choice, and thus, benefit to that appreciation for the magnitude and variability of participant; and (c) the ease of quantifying other the costs of participating in the interactions. In so variables relevant to the ant’s partner (e.g. defolia- far as the benefit accrued by one partner translates tion, aphid mortality). Ant-fungal mutualisms are into the cost experienced by the other, conflicts of an exception; microbe partners or substrates can be interest between ants and their prospective mutu- manipulated, and outcomes for colony growth or alists may seem unavoidable (e.g. Section 6.2.5 and survival can be measured relatively easily (e.g. Fer- Figure 6.4, see also Palmer et al. 2008; Stanton et al. na´ndez-Marı´n et al. 2007; Seal and Tschinkel 2007a). 1999; Yu and Pierce 1998). However, three largely One solution to this widespread shortcoming is to untested hypotheses explain why this need not be utilize the modular organization of ant colonies, the case: and to more fully describe the effects of these inter- The resources or strategies that benefit one participant actions on individual modules (i.e. individual ants, may come at negligible cost to its partner. In such see Cushman et al. 1994; Lach et al. 2009). Stable situations, the benefit to the recipient may not isotope techniques are one new promising tech- come at a commensurate cost to the provider, and nique to quantify the benefits that ants receive one can get ‘something for nothing’. The costs of (e.g. Sagers et al. 2000; Box 7.1). Davidson et al. plant-produced rewards can be quite minor (2003) used stable isotopes to infer that access to (O’Dowd 1979, 1980), and are lessened when plants extrafloral nectar and hemipteran exudates in the decrease or curtail extrafloral nectar production in rainforest canopy fuel the spectacular diversity and the absence of perceived threats (Lach et al. 2009; abundance of ants in those habitats. Critically, sta- Ness 2003a). Indeed, the multiple prospective ben- ble isotope techniques highlight the integration of efits of ant attendance to hemipterans beg the ques- rewards into ant tissues or particular castes rather tion, why do not all hemipteran species exchange than measuring fitness, and their correct interpreta- their waste products for ant attendance? Similarly, tion requires a comprehensive knowledge of the some of the benefits ants provide likely incur negli- natural history of the system. gible cost to the colony. For example, some plants Barring obligate ant–myrmecophyte or ant–fun- benefit from access to the debris accumulated by gal interactions, we know of no studies that have foraging ants and the CO2 they exhale (e.g. Sagers sought evidence (much less demonstrated) that et al. 2000; Treseder et al. 1995; Wagner 1997). Last, among-site variation in ant communities is attribut- although participation in particular mutualisms able to variation in the availability of their mutual- may incur costs, those solutions are often cheaper ists (but see Dı´az-Castelazo et al. 2004; O’Dowd et al. than the alternatives (e.g. myrmecochory versus 2003). The inherent asymmetry in many ant–mutu- frugivory in nutrient-poor habitats; Westoby et al. alist interactions offers one explanation for that ab- 1991b). sence. Ants often provide their plant and The resources traded in these interactions may be trophobiont partners with protection at critical de- less important than are other components of the in- mographic stages (e.g. protection of seeds and teraction. For example, many ant-collected seeds 112 ANT ECOLOGY include odorants that elicit collection behaviours play a central role in shaping the ecology and evo- by workers (e.g. Hughes et al. 1994), and in lution of ant–mutualist interactions. For example, it some cases an elaiosome reward is entirely ab- has been proposed that EFNs have evolved as a sent (e.g. Youngsteadt et al.2008).Insomere- means of distracting ants from tending hemipterans spects, these interactions may function more as (Becerra and Venable 1989) and collecting floral ‘behavioural usurpation’ than a reciprocal ex- nectar (Wagner and Kay 2002). However, ant- change of resources. Ecologists have yet to ask tended insects are disproportionately common on whether diaspores will be collected when these EFN-bearing plants (Offenberg 2000), and some compounds are experimentally disassociated even ingest extrafloral nectar (DeVries and Baker with the seed, although we know that other 1989). The rewards provided by hemipterans can non-rewarding substances impregnated with the also supplement the ant rewards provided by myr- volatiles will be collected by workers (e.g. mecophytes (Fonseca 1993; Palmer et al. 2008). In so Hughes et al. 1994). far as ants have greater control over hemipteran The outcome that benefits the myrmecophile may be a densities than they do direct plant rewards, the productofstrategiesthatbestsuittheant. In such a involvement of these third parties can affect the case, the concept of ‘cheating’ becomes meaning- functioning of the symbiosis (Gaume et al. 1998). less. For example, if an ant colony is capable of the Some plants also utilize ants within the context of vigorous defence of a resource against real or per- multiple mutualisms involving protection and seed ceived competitors, be they rival colonies, herbi- dispersal (Turnera ulmifolia: Cuatle et al. 2005; Urera vores, or carnivores, it will do so. If it cannot, the baccifera: Dutra et al. 2006). The most thoroughly opportunity to harvest that resource may well be studied of these systems, and perhaps the most usurped by a more aggressive colony that pro- reticulate, is Calathea ovandensis; this tropical herb vides even greater protection to the reward (plant has EFNs, is attacked by ant-tended Lepidoptera, or insect). On a different vein, Ness et al. (2009) and relies on myrmecochory for seed dispersal demonstrated that sustained collection of carbohy- (Horvitz and Schemske 1984, 1986). How often the drate-rich resources changes ant dietary prefer- coterie that participates in one interaction is well ence, and inferred that an abundance of one suited for the other, or interacts with that counter- resource highlights the relative absence of comple- part, is unknown (but see Cuatle et al. 2005). Ex- mentary resources (here, provision of abundant plorations of these inter-mutualism dynamics may carbohydrates elicit attacks on relatively nitro- provide important insight into the evolution of ant gen-rich prey). Last, from a myrmecochore’s per- mutualisms generally. spective, the most important characteristic of an The multiple mutualisms and complex interac- ant is that it does not ‘cheat’ by removing the tions occurring within the nests of fungal garden- elaiosome and abandoning the denuded seed ing ants are a rich area for exploring potentially (to predators, competitors, etc). For subordinate competing mutualisms and the effects of parasites. ants that specialize in discovering but not domi- The mutualism between actinomycete bacteria and nating resources, the most advantageous beha- ants and the parasitic black yeast-actinomycete viour may be to immediately collect the bacteria and Escovopsis–fungal cultivar relation- elaiosome (with the seed attached) rather than ships have only recently been discovered to sub- engaging in the time-consuming task of separating stantially shape the dynamics of the ant–fungal reward from seed, and hence risk losing the re- cultivar mutualism. Questions remain about the source to a competitor. mechanisms through which some effects are seen. For example, given that actinomycete bacteria are stimulated by the presence of Escovopsis (Currie et 6.8.4 Inter-mutualism conflict al. 2003), are black yeasts as well? And if so, are Relatively little is known of the interactions be- there feedback mechanisms by which black yeasts tween mutualisms, including those in which ants facilitate Escovopsis infection? It is likely that are involved. Such inter-mutualism dynamics may new microbes that may further affect the costs ANTS AS MUTUALISTS 113 and benefits of the multiguild relationships await these interactions rarely occur in proportion to the discovery. density or diversity of these interactions. How might our understanding of these interactions change if we studied them in the settings where 6.8.5 Biotic interactions on an abiotic stage they most often occur? Most studies of myrmecoch- Experimental studies have begun to explore the ory (and all that quantify benefit to the ants) focus importance of variation in abiotic resources on the on temperate deciduous myrmecochores; what do evolution and functioning of ant mutualisms. In so the costs and benefits described in this nutrient-, far as these resources are limiting, they can alter the moisture-, and granivore-rich biome tell us about incentives for particular interactions. For example, the >90% of myrmecochores that reside in dissimi- carbon-rich resources such as extrafloral nectar and lar biomes in Australia and South Africa? Might our ant domatia should be less costly for plants to pro- sense of the costs, benefits, and selection pressures duce where carbon is in excess (Folgarait and on ant-tended insects and plants differ if we stud- Davidson 1994). Perhaps as a result, EFN-bearing ied them in communities such as tropical rainforest plants are common in sunlight-rich habitats such as canopies or some deserts where tending by ants is rainforest canopies (Blu¨ thgen et al. 2000), forest the modal interaction? The characteristics of ants, edges (Bentley 1976), and deserts (Pemberton prospective natural enemies, and competition for 1988). The influence of resource limitation (or sur- services may be sufficiently different in these set- plus) is also detectable at smaller spatio-temporal tings to profoundly alter those interactions. We scales. Nitrogen fertilization of host plants can in- recognize that particular systems offer advantages crease tending rates of some trophobionts (e.g. ly- for studying particular ecological and/or evolu- caenids: Billick et al. 2005, but see Morales and Beal tionary phenomena. However, we propose that 2006 re. membracids), and alter plant investment in the wealth of studies that comprise the current lit- indirect defences (Folgarait and Davidson 1995). erature and inform reviews, meta-analyses, and our Ant mutualisms can also reorganize abiotic re- gestalt sense of how interactions function may de- sources. The construction, maintenance, and feed- scribe the range and modal version of the interac- ing of ant colonies and nests often concentrate tions only in so far as those studies occur in resources, expose buried nutrients, and alter mois- comparable environments. Our understanding of ture retention rates (Moutinho et al. 2003), perhaps those interactions will change as we better place to the benefit of their partners (Davidson and Mor- our questions in the context of the larger environ- ton 1981; Giladi 2006; Wagner 1997). Remarkably, ment. there is also evidence that tending by ants can alter the nitrogen content of tended hemipterans and 6.9 Summary their host plants (Kay et al. 2004, but see Abbot et al. 2008). The generality of these documented pat- Ants are perhaps the most common and dominant terns, and how such modifications will influence animal mutualists in terrestrial environments. As a the incentives for ant mutualisms, is largely un- result, better understanding the dynamics of these known. To make matters more complex (and wor- interactions should be a priority for those who hope thy of attention), the availability of nitrogen and to understand the taxon, their role in communities,

CO2 continues to increase at scales ranging from and mutualism as a widespread interspecific interac- individual plants to the biosphere as a result of tion. These mutualisms include interactions with ant- anthropogenic influences. loving plants, insects, fungi, and bacteria, with the ants typically receiving food and/or shelter, and their partners receiving food, protection, and/or 6.8.6 Putting ant mutualisms in their place propagule dispersal. Context dependency, wherein Ant mutualisms are unevenly distributed across the magnitude of costs and benefits incurred as a habitats. Perhaps problematically, the settings for result of participation in the interactions varies with the research that underpins our understanding of the ecological setting may be particularly prevalent 114 ANT ECOLOGY in interactions involving protection by ants. Adapta- costs and benefits to participants, and addresses how tions that enable effective participation in one type of contemporary interactions and abiotic resources alter mutualistic interaction may preclude a species from these interactions. participating in others, but may also make it more difficult for non-beneficial interactors to intrude. Acknowledgments Because ant interactions with their mutualists are relatively easily monitored, manipulated, and are This work was greatly improved by the comments sustained over time, ant mutualisms are model of Aaron Gove, Paulo Oliveira and Kirsti Abbott. systems for understanding mutualisms and plant Joshua Ness was supported by sabbatical funds defence. We encourage future work that explores from Skidmore College during the composition of the influence of partner diversity, better quantifies this chapter.